# Small language model

> A compact language model that is cheap to run, even on a laptop or a phone.

- Level: Intermediate
- Topics: AI
- URL: https://tidbit.legios.com.ar/small-language-model/
- In Spanish: [Modelo de lenguaje pequeño](https://tidbit.legios.com.ar/es/small-language-model/)

A **small language model** (SLM) is a language model with far fewer parameters than the giant ones behind the best-known chatbots. There is no official cutoff, but the term usually means models from a few hundred million to roughly ten billion parameters, against hundreds of billions or more for the frontier models. Examples include Microsoft's Phi family, Google's Gemma and the smaller versions of Meta's Llama.

Smaller means cheaper and faster. An SLM can run on a single modest GPU, on a laptop, or directly on a phone, with no internet connection. That is useful for privacy (the text never leaves the device), for latency (no round trip to a data center) and for cost when you have to process millions of short requests, like tagging support tickets or pulling fields out of invoices.

The catch is breadth. A small model knows less about the world, gets lost more easily in long, multi-step tasks and hallucinates more on topics it barely saw. Teams often get around this by fine-tuning a small model on one narrow job, where it can match a much bigger one, or by training it to imitate a large model (a technique called distillation). A common setup is a mix: a small model handles the easy, frequent work and passes the hard cases to a large one.

## An analogy

A pocket dictionary next to an encyclopedia. It will not answer everything, but it fits in your bag, opens instantly and is enough for most of what you look up every day.

## Keep in mind

- No official size cutoff; usually up to roughly ten billion parameters.
- Can run on laptops and phones, even offline.
- Cheaper, faster and more private, but with less general knowledge.
- Shines when fine-tuned for one narrow, well-defined task.

## Go deeper

- [What are small language models? (IBM)](https://www.ibm.com/think/topics/small-language-models)
- [Small language model on Wikipedia](https://en.wikipedia.org/wiki/Small_language_model)

## Related terms

- [LLM](https://tidbit.legios.com.ar/llm/): A neural network trained on huge amounts of text to predict the next word.
- [Chatbot](https://tidbit.legios.com.ar/chatbot/): A program you talk to by typing or speaking, as if it were a person.
- [Microsoft](https://tidbit.legios.com.ar/microsoft/): The company of Windows and Office, now just as much a cloud company with Azure.
- [Google](https://tidbit.legios.com.ar/google/): The search company that gives most products away and makes its money on ads.
- [Meta](https://tidbit.legios.com.ar/meta/): The company behind Facebook, Instagram and WhatsApp, paid for by advertising.
- [GPU](https://tidbit.legios.com.ar/gpu/): The chip that does thousands of simple calculations at once.
- [Laptop](https://tidbit.legios.com.ar/laptop/): A whole computer folded into one portable case that runs on a battery.
- [Internet](https://tidbit.legios.com.ar/internet/): A worldwide network of networks that lets any connected device reach any other.